top of page

Seamless Integration

preprocessing
soft voting diagnosis
data creation
copilot logo
copilot logo
copilot logo

Reteena Copolit is able to accurately super resolve and segment low-field MRI scans to produce enhanced scans

scan enhancement

Image Enhancement Model Metrics

Model trained on simulated scans which may make it less accurate for real low-field MRI scans

model performance

Diagnosis
Model Metrics

Bias toward certain age range and may be inaccurate for users not in that age range

model performance overall

Scan Acquisition

600 deidentified expertly segmented 3D High Field (HF) MRI scans were obtained from 200+ patients ranging from the ages of 50-80 from the OASIS 3 Brains Dataset.

600

200+

50 -80

Scans were taken on T1w and T2w scanners with Tesla strengths of either 1.5T and 3T from various hospitals and diagnosis centers around the United States. These scans were auatomatically segmented for various brain regions using FreeSurfer software before being validated and modified by a radiologist to increase the accuracy of the segmentations. Each patient within the dataset was also diagnosed for AD by a specialist and underwent multiple MR sessions.

Scan Preprocessing

Raw scans were pre processed before being passed into my deep learning framework for image enhancement in order to increase computational efficiency. This is because the raw scans have high dimensionality (up to 256 x 256 x 256) and a large array of varying pixel intensities. A preprocessing pipeline is necessary to reduce computational strain and remove unnecessary information within scans.

256 x 256 x 256

Reteena

Reteena is fiscally sponsored by The Hack Foundation (d.b.a. Hack Club), a 501(c)(3) nonprofit (EIN: 81-2908499).

© 2024 Reteena. All rights reserved.

About

Contact

Research

Support

bottom of page